Guolian Minsheng Securities: Commercialization of AI applications fully initiated, reconstructing intelligent terminal ecosystem

Guolian Minsheng Securities released a research report stating that China and the United States are ushering in a new wave of large-scale model innovation, and China's global advantage in AI applications comes from the synergy of three capabilities, making its barriers difficult to replicate. On the demand side, the global leader in AI computing power is transitioning from training to inference. On the supply side, the construction of super nodes, software ecosystems, and other factors will lay an important foundation for the long-term rise of domestic computing power. On one hand, the breakthrough in large model capabilities unlocks the technological bottleneck of AI applications, and on the other hand, competition among giants for AI super portals will drive AI applications into thousands of industries and millions of households. The main points of Guolian Minsheng Securities are as follows: - The commercialization of AI applications is fully opening up, and a new wave of large model innovation is coming to China and the United States. - Gemini 3 has lifted the limit on large models, making AI applications transition from "available" to "usable," while Google has already established a mature "computing model-application" internal loop as a blueprint for global AI industry development. The development of the AI industry has accelerated the commercialization of AI applications globally. Chinese AI applications are seeing rapid growth in overseas revenue, active users, and other operational metrics. According to extraordinary research data, AI applications represented by Kaling, PLAUD, HeyGen, etc., have achieved nearly $100 million in ARR overseas, covering multiple tracks such as video generation, article summarization, intelligent bodies, etc. China's global advantage in AI applications comes from the synergy of three capabilities, making its barriers difficult to replicate. - AI is reshaping the intelligent terminal ecosystem, opening up a "Three Kingdoms" situation. - Large model manufacturers are shifting from upgrading model capabilities to penetrating consumer scenarios, while the model calling ecosystem is laying a solid foundation for consumer applications to land; terminal hardware manufacturers are grasping the autonomy of operating systems to develop native AIOS, reshaping the core logic of AI phones; super application ecosystems occupy the largest C-end traffic entry points to maximize user value. Hundreds of MAUs of traffic + high-value user data + vertical scene know-how are expected to become the core barriers for C-end manufacturers. - The first year application of domestic AI computing power, from training to inference computing tipping point. - On the demand side, the global AI computing power leader is transitioning from training to inference. Factors such as the improvement of domestic large model inference efficiency and the rapid growth of core application users will drive AI computing power towards the tipping point of inference. On the supply side, the construction of super nodes, software ecosystems, and other factors will lay an important foundation for the long-term rise of domestic computing power. - Commercial spaceflight and quantum technology: Facing the new productivity of "galaxies and oceans." - In the strategic stage of domestic low-orbit satellite networking, important breakthroughs in the launch and recovery end are of significance, and the space-time computing power and other areas are expected to benefit in the long term. Quantum technology, as a cutting-edge technology, is expected to become an important engine for the development of new quality productivity. Risk warning: AI industry development falls short of expectations, technology industry policies are not implemented as expected, and industry competition intensifies.